طراحي تراشه بينايي تشخيص لبه براي كاربردهاي بلادرنگ بر اساس مدل بينايي موجودات زنده

چكيده انگليسي :

Designing an Edge Detector Vision Chip for Real Time Applications Based on Biological Vision Systems Hesam Ahmadi h ahmadi@ec iut ac ir December 28 2012 Department of Electrical and Computer Engineering Isfahan University of Technology 84156 83111 IranDegree M Sc Language FarsiSupervisor Dr Sayed Masoud Sayedi m sayedi@cc iut ac irAbstractEdge detection is a demanding need in robotic military and automation areas and is used as an early image processingin high level image processors such as movement detectors target tracker and etc A typical technique for edgedetection is the use of sensor processor combination In this combination raw images data are transferred to processorthrough a data line This creates a bottleneck that reduces process speed Due to the existence of this bottleneck running real time software edge detection requires very high speed processors that tend to be power consuming Anattractive solution to this problem is to shift part of the image data processing into the sensor hardware structure It isdone by designing and allocating some proper circuitry around image sensors to perform some early pixel level imageprocessing within the photo sensor array Intelligent vision chips as a new generation of image processing systems bycombining image sensor and processing circuits in a single chip provide the opportunity for parallel data processingwith higher speed and lower power consumption Also due to the integration of these systems area consumption willbe reduced In this work a kind of vision chip that has the ability of real time edge detection has been designed andsimulated in CMOS 180nm technology The chip can recognize color and intensity edges by comparing color andintensity data of each pixel with neighboring pixels It produces a two bit result for each pixel which represents thetype and the existence or absence of the edge in that pixel and hence provides the next processing stages with lessrequirements in terms of computational power cost portability and power consummation To implement thesystem the mechanisms of biological vision systems in color intensity and edge detections are studied and theproposed detector has been designed based on retina models in color and edge detections Buried Triple Junctionphoto detector BTJ is used as the pixel light sensing element Color and intensity detections are done with only oneBTJ by using the color filtering feature of silicon The sensor determines three components of color that are used forboth color and intensity detections With this technique the system is compact the power requirement is reduced andthe chip is implementable in a standard CMOS technology with no extra processes Light input dynamic range hasbeen improved by using logarithmic amplifiers in the input stage of each pixel The designed system has 80um 60umpixels consumes 135nw power per pixel and works with 1v power supply It can be used in portable systems Systemfunctionality has been verified through simulations The results show that system can process image streams with upto 30 frames per second speed in real time Also they show that intensity edge detector has only high sensitivity tolight intensity and has low sensitivity to light color and typically detects only intensity edges Similarly simulationresults show that color edge detector has good sensitivity to color changes and detects color edges if only one colorcomponent changes in a pixel place The physical layout level of the design and post layout simulations have beenpresented in CMOS 180nm technology Keywordsedge detection intelligent vision chips biological vision systems color detection low powercircuits

احمدي، حسام

طراحي تراشه بينايي تشخيص لبه براي كاربردهاي بلادرنگ بر اساس مدل بينايي موجودات زنده

تشخيص رنگ , مدارهاي با توان پايين